Han et al. (2026) Spatiotemporal evolution and future trends in vegetation productivity in Xinjiang driven by vapor pressure deficit and potential evapotranspiration

Identification

• Journal: Ecological Indicators
• Year: 2026
• Date: 2026-09-25
• Authors: Yongjin Han, Hengfang Wang, Hao Huang
• DOI: 10.1016/j.ecolind.2026.115553

Research Groups

• College of Ecology and Environment, Xinjiang University, Urumqi, China
• Key Laboratory of Oasis Ecology of the Ministry of Education, Xinjiang University, Urumqi, Xinjiang, China
• Ministry of Education Xinjiang Jinghe Temperate Desert Ecosystem Observation and Research Station, Bortala Mongolian Autonomous Prefecture, Xinjiang, China

Short Summary

This study investigated the spatiotemporal evolution and future trends of vegetation productivity in Xinjiang, China, from 2004 to 2023, focusing on the regulatory mechanisms of atmospheric drought indicators (vapor pressure deficit and potential evapotranspiration). The findings reveal spatially heterogeneous responses, with overall GPP projected to decline by 2033, highlighting the need for differentiated regional management strategies.

Objective

• To analyze the spatial patterns of temporal changes in vegetation productivity in Xinjiang from 2004 to 2023.
• To determine how the responses of gross primary productivity (GPP) and normalized difference vegetation index (NDVI) to drought stress indicators, including vapor pressure deficit (VPD) and potential evapotranspiration (PET), vary spatially across Xinjiang.
• To project the potential future trends in vegetation productivity in Xinjiang under climate change.

Study Configuration

• Spatial Scale: Xinjiang Uygur Autonomous Region, China (approximately 1.6649 × 10^6 km^2), analyzed at a 1 km pixel resolution.
• Temporal Scale: Historical analysis from 2004 to 2023 (20 years); Future projection from 2024 to 2033 (10 years). Data aggregated to an annual scale.

Methodology and Data

• Models used:
  • Trend analysis: Theil–Sen median slope estimation, Mann–Kendall (MK) nonparametric test.
  • Correlation analysis: Pearson correlation, Partial correlation.
  • Feature importance/interpretability: SHapley Additive exPlanations (SHAP) with eXtreme Gradient Boosting (XGBoost) regression model.
  • Persistence analysis: Hurst exponent.
  • Future trend prediction: Ordinary Least Squares (OLS) regression, Random Forest (RF), XGBoost.
  • Statistical tests: Augmented Dickey–Fuller (ADF) stationarity test, Diebold–Mariano (DM) test.
• Data sources:
  • Vegetation productivity/greenness: NASA's Moderate Resolution Imaging Spectroradiometer (MODIS) products (NDVI: MOD13A1, GPP: MYD17A2H, FPAR: MOD15A2H) from AppEEARS, with 500 m spatial resolution.
  • Climate data: National Geoscience Data Center for annual precipitation (Pre), annual mean temperature (TMP), and potential evapotranspiration (PET), with 1 km spatial resolution.
  • Vapor pressure deficit (VPD) data: Copernicus Climate Data Store, with 0.01° spatial resolution.

Main Results

• Vegetation productivity in Xinjiang showed pronounced spatial heterogeneity from 2004 to 2023. GPP significantly increased in 22% of the study area, primarily along the Tarim Basin margins and southern Tianshan Mountains, while NDVI significantly improved in 8.9% of the region, centered on southern Xinjiang's oasis belts. Localized vegetation degradation was observed in northern and central Xinjiang.
• VPD and PET were identified as the most important explanatory variables for vegetation productivity variation, with substantially higher relative importance than temperature and precipitation.
• Temporally, both VPD and PET showed weak but significant negative correlations with NDVI (r = -0.27 and r = -0.28, respectively) and GPP (r = -0.21 and r = -0.22, respectively).
• Spatially, the relationships between drought indicators and vegetation productivity were heterogeneous: positive correlations were found in southern oasis regions, while negative correlations predominated in northern Xinjiang. This spatial differentiation remained significant after controlling for temperature and precipitation.
• Hurst exponent analysis indicated strong long-term persistence in GPP and NDVI trajectories in only 8.6% of vegetated areas, mainly in the southern Tianshan Mountains and northern Tarim Basin.
• Multi-model simulations projected a decline in mean GPP to approximately 197.62 g C m−2 yr−1 by 2033, after peaking in 2019, under the scenario of continued historical trends of increasing VPD and PET.
• The XGBoost model demonstrated high predictive accuracy for GPP (R^2 = 0.78) and NDVI (R^2 = 0.59), outperforming the Random Forest model.

Contributions

• This study provides a comprehensive spatiotemporal analysis of vegetation productivity dynamics in Xinjiang, China, under atmospheric drought stress, using a multi-model approach including advanced machine learning and persistence analysis.
• It quantitatively demonstrates that vapor pressure deficit (VPD) and potential evapotranspiration (PET) are primary drivers of vegetation productivity, with their relative importance exceeding other climatic factors.
• The research reveals a critical spatial dichotomy in the response of vegetation productivity to atmospheric drought: positive correlations in southern oasis regions (likely due to irrigation) versus negative correlations in northern Xinjiang, a key contribution that moves beyond average responses.
• The study projects a future decline in gross primary productivity (GPP) by 2033, highlighting the increasing risk of vegetation degradation under continued climate change.
• It underscores the necessity for differentiated, region-specific ecological management strategies in Xinjiang, advocating for optimized water allocation in southern oases and enhanced drought monitoring and active restoration in northern degraded areas.

Funding

• National Natural Science Foundation of China (32360282)
• Key Research and Development Project of Xinjiang Uygur Autonomous Region (2024B03017)
• Xinjiang Uygur Autonomous Region's Tianshan Talent Training Program (2024TSYCCX0017)

Citation

@article{Han2026Spatiotemporal,
  author = {Han, Yongjin and Wang, Hengfang and Huang, Hao},
  title = {Spatiotemporal evolution and future trends in vegetation productivity in Xinjiang driven by vapor pressure deficit and potential evapotranspiration},
  journal = {Ecological Indicators},
  year = {2026},
  doi = {10.1016/j.ecolind.2026.115553},
  url = {https://doi.org/10.1016/j.ecolind.2026.115553}
}